JackFramework

How to build on, launch, and debug JackFramework (JF) — a PyTorch training framework that owns the training loop, DDP launcher, checkpoint I/O and progress bar while you supply a model-interface class, a dataloader class and a launcher. Use this whenever a task touches JF hooks (inference / loss / accuracy / post_process / get_train_dataset / save_result), launching `python Source/main.py --mode train|test`, JF CLI flags (--modelDir --modelName --dataset --dist --gpu --imgNum --dataloaderType), checkpoints and checkpoint.list, warm-starting from a prior .pth, or any JF run that failed. Trigger it even when JF is never named — e.g. "training silently skips my loss function", "dies with an empty ProcessRaisedException", "the progress bar never reaches 100%", "checkpoint list file not found", "the eval finished but there's no output csv", "how do I resume from epoch 4". Most JF failures are silent, or put the traceback nowhere near the real cause, so consult this before guessing at a fix.